The Agent Skills Directory

Data Ingestion Surface: The skill defines workflows in SKILL.md for processing text from the dbo.documents table. This ingestion point for untrusted data represents a potential surface for indirect prompt injection. The documentation does not specify boundary markers or sanitization routines for the processed content. The skill's capability inventory includes network-facing operations like AI_GENERATE_EMBEDDINGS and sp_invoke_external_rest_endpoint, which process this data.
Network Interaction Pattern: The skill utilizes sp_invoke_external_rest_endpoint to reach domains such as openai.azure.com. These are official Azure service endpoints managed by the vendor. These network operations are constrained by database-scoped credentials and a platform allowlist, which are standard security controls for this environment.
Database Permission Management: The guide includes SQL GRANT statements for permissions like EXECUTE ON EXTERNAL MODEL. These instructions are intended to configure access for application users within the database's security model, facilitating the use of the described features.

embeddings-and-external-models